vc4: Optimize CL emits by doing size checks up front.

The optimizer obviously doesn't have the ability to rewrite these to skip
the size checks per call, so we have to do it manually.

Improves a norast benchmark on simulation by 0.779706% +/- 0.405838%
(n=6087).

1 files changed, 8 insertions, 4 deletions

diff --git a/src/gallium/drivers/vc4/vc4_cl.c b/src/gallium/drivers/vc4/vc4_cl.c
index 36dd28c..0700e88 100644
--- a/src/gallium/drivers/vc4/vc4_cl.c
+++ b/src/gallium/drivers/vc4/vc4_cl.c
@@ -29,17 +29,21 @@ void
 vc4_init_cl(struct vc4_context *vc4, struct vc4_cl *cl)
 {
         cl->base = ralloc_size(vc4, 1);
-        cl->end = cl->next = cl->base;
+        cl->next = cl->base;
+        cl->size = 0;
 }
 
 void
-vc4_grow_cl(struct vc4_cl *cl)
+cl_ensure_space(struct vc4_cl *cl, uint32_t space)
 {
-        uint32_t size = MAX2((cl->end - cl->base) * 2, 4096);
+        if ((cl->next - cl->base) + space <= cl->size)
+                return;
+
+        uint32_t size = MAX2(cl->size + space, cl->size * 2);
         uint32_t offset = cl->next -cl->base;
 
         cl->base = reralloc(ralloc_parent(cl->base), cl->base, uint8_t, size);
-        cl->end = cl->base + size;
+        cl->size = size;
         cl->next = cl->base + offset;
 }